The revolution represented by third-generation photovoltaic devices relied on the discovery of various hybrid organic–inorganic perovskite materials to convert solar into electrical energy. One of the advantages of such cells is their low cost due to the raw materials and cheap production methods used. Nevertheless, these cells face several challenges, such as inadequate stability and the hysteresis phenomenon. To overcome these, perovskite solar cell (PSCs) with planar and inverted structures have been utilized with an inorganic hole transport layer (HTL), achieving acceptable efficiency. As there is no closed-form system of equations to describe the operation of such cells, neural networks have been employed for their modeling. In optimization algorithms, the values of the parameters must be swept, since most current simulation tools cannot use them directly. Such software optimization can notably decrease the cost of cell design. This paper presents a practical way to achieve the mentioned aim. In particular, an artificial neural network (ANN) is exploited for the modeling, then an evolutionary particle swarm optimization (E-PSO) algorithm is developed to optimize the structure to achieve the highest efficiency based on searching the energy conversion. The results of the simulations are then employed in SCAPS software to train the neural network. This optimization leads to the achievement of an efficiency of 23.76% for the proposed structure, better than values reported in literature.

以第三代光伏设备为代表的革命依赖于发现各种有机-无机钙钛矿杂化材料，将太阳能转化为电能。这种电池的优点之一是由于使用的原材料和便宜的生产方法而使其成本低廉。然而，这些电池面临一些挑战，例如稳定性不足和滞后现象。为了克服这些问题，具有平面和倒置结构的钙钛矿太阳能电池（PSC）已与无机空穴传输层（HTL）一起使用，从而获得了可接受的效率。由于没有封闭的方程式系统来描述此类细胞的运行，因此已将神经网络用于其建模。在优化算法中，必须清除参数值，因为大多数当前的仿真工具无法直接使用它们。这种软件优化可以显着降低单元设计的成本。本文提出了一种实现上述目标的实用方法。尤其是，利用人工神经网络（ANN）进行建模，然后开发了进化粒子群优化（E-PSO）算法，以在搜索能量转换的基础上优化结构以实现最高效率。然后将仿真结果用于SCAPS软件中以训练神经网络。通过优化，该结构达到了23.76％的效率，优于文献报道的值。尤其是，利用人工神经网络（ANN）进行建模，然后开发了进化粒子群优化（E-PSO）算法，以在搜索能量转换的基础上优化结构以实现最高效率。然后将仿真结果用于SCAPS软件中以训练神经网络。通过优化，该结构达到了23.76％的效率，优于文献报道的值。尤其是，利用人工神经网络（ANN）进行建模，然后开发了进化粒子群优化（E-PSO）算法，以在搜索能量转换的基础上优化结构以实现最高效率。然后将仿真结果用于SCAPS软件中以训练神经网络。通过优化，该结构达到了23.76％的效率，优于文献报道的值。